张秋昭,何心月,张龙强,等.基于高位远程滑坡降雨入渗数值模拟的GNSS/加速度计融合变形监测与预警分析[J].河南理工大学学报（自然科学版）,2026,45(6):80-90.

Zhang Q Z, He X Y, Zhang L Q, et al.Deformation monitoring and early warning analysis of high-position long-runout landslides based on GNSS/accelerometer fusion and rainfall infiltration numerical simulation[J].Journal of Henan Polytechnic University(Natural Science) ,2026,45(6):80-90.

基于高位远程滑坡降雨入渗数值模拟的GNSS/加速度计融合变形监测与预警分析

张秋昭1,2,3, 何心月1,2, 张龙强1,2, 刘鑫1,2,3, 郑南山1,2,3

1.中国矿业大学 露天煤矿灾害防治与生态保护全国重点实验室，江苏 徐州  221116；2.中国矿业大学 环境与测绘学院，江苏 徐州  221116；3.中国矿业大学 自然资源部国土环境与灾害监测重点实验室，江苏 徐州  221116

摘要: 目的 面向高寒高原关键生态走廊中以降雨入渗为触发机制的高位远程滑坡，针对其“高差大、滑程长、加速段短促”的运动特征导致的监测与预警难题，采用数值模拟生成的仿真观测数据开展分析，提出一种融合GNSS/加速度计多速率卡尔曼滤波算法与速率阈值判据的分级预警方法。  方法 基于高寒高原关键生态走廊区域滑坡特征的多物理场数值模拟生成的仿真数据，利用COMSOL Multiphysics平台建立固体力学与理查兹方程耦合的“表面渗流-岩体损伤”二维有限元模型，得到多监测点位移/加速度的参考真值时间序列。在此基础上，叠加观测噪声与缓变漂移构建GNSS与加速度计的仿真观测；融合环节采用多速率卡尔曼滤波：以高频加速度驱动状态预测、低频GNSS执行量测更新，在采样比约束下实现联合估计；构建基于速率阈值的分级预警方法。  结果 在仿真数据下融合方案相较单GNSS在关键指标上表现更优：监测精度RMSE相对提升37.4%、位移轨迹更为平滑、预警触发时刻与滑坡模型加速转折更一致。 结论 以“数值模拟、仿真观测、多速率卡尔曼融合、预警分析”为核心的系统评估表明，该方法可在降雨入渗型高位远程滑坡失稳前的中后期加速阶段更及时且稳定地维持高等级警戒，旨在为高寒高原关键生态走廊区域滑坡现场监测、预警参数确定提供参考。

关键词:高位远程滑坡;GNSS;加速度计;数据融合;监测预警

doi:10.16186/j.cnki.1673-9787.2025100011

基金项目:国家自然科学基金资助项目（U22A20569，42304046）

收稿日期:2025/10/09

修回日期:2026/01/27

出版日期:2026/10/09

Deformation monitoring and early warning analysis of high-position long-runout landslides based on GNSS/accelerometer fusion and rainfall infiltration numerical simulation

Zhang Qiuzhao1,2,3, He Xinyue1,2, Zhang Longqiang1,2, Liu Xin1,2,3, Zheng Nanshan1,2,3

1.State Key Laboratory of Disaster Prevention and Ecology Protection in Open-pit Coal Mines， China University of Mining and Technology， Xuzhou  221116， Jiangsu， China；2.School of Environment and Spatial Informatics， China University of Mining and Technology， Xuzhou  221116， Jiangsu， China；3.Key Laboratory of Land Environment and Disaster Monitoring， China University of Mining and Technology， Xuzhou  221116， Jiangsu， China

Abstract: Objectives To address the monitoring and early warning challenges posed by high-position， long-runout landslides triggered by rainfall infiltration in key ecological corridors of cold and high-altitude regions—characterized by large elevation differences， long sliding distances， and short acceleration phases—a graded early warning method that integrates a GNSS/accelerometer multi-rate Kalman filter with velocity threshold criteria is proposed， based on synthetic observational data generated through numerical simulation. Methods Using the COMSOL Multiphysics platform， a two-dimensional finite element model coupling solid mechanics with the Richards equation was developed to simulate surface seepage and rock damage， generating reference time series of displacement and acceleration at multiple monitoring points. Synthetic GNSS and accelerometer observations were then constructed by adding noise and slow drift to the reference data. A multi-rate Kalman filter was applied for data fusion： high-rate acceleration data drive the state prediction， while low-rate GNSS data perform the measurement update， enabling joint estimation under sampling-ratio constraints. A velocity-threshold-based graded early warning scheme was subsequently established.  Results Under simulated conditions， the fusion scheme outperforms GNSS-only monitoring in key metrics： RMSE improves by 37.4%， displacement trajectories become smoother， and warning trigger times align more closely with the model’s acceleration turning points.  Conclusions The systematic evaluation—encompassing numerical simulation， synthetic observation， multi-rate Kalman fusion， and early warning analysis—demonstrates that the proposed method can maintain timely and stable high-level alerts during the middle-to-late acceleration stage before failure of rainfall-induced high-position long-runout landslides. This study aims to provide a reference for field monitoring and warning parameter determination for landslides in cold and high-altitude ecological corridors.

Key words: high-position long-runout landslide; GNSS; accelerometer; data fusion; deformation monitoring and early warning